
@LauraVickersTM Skrocone do 1/3 byłoby ciekawsze. ChatGPT jest bardzo rozgadany i lubi powtarzac wszystko po 3 razy.
A teza tego artykułu jest fałszywa, kod to tylko niewielka część wartościowych danych.
Polski
Dariusz Swierk PhD
1.7K posts

@swierk
Theoretical physics: a lot of math, imagination and Python. Editor in Chief: https://t.co/G4RlWL1LaJ Book: "AI: Dark Scenarios"














Crazy idea: What if Europe built a niche open-source model focused only on coding? Just the best coding model we can build. Open weights. European talent. Community-driven. Would anyone be interested in contributing to an open-source project like this?







Sam Altman calls Yann LeCun’s bet against LLM scaling as “misguided”. “So clearly LLMs are capable of figuring out new knowledge and clearly they are capable of doing some things that humans just can't do. they are going to scale much further.”
















